1. Field of the Invention
The present invention relates generally to surface emitting lasers, and particularly to tunnel junctions for long-wavelength vertical cavity surface emitting lasers (VCSELs).
2. Technical Background
Vertical cavity surface emitting lasers (VCSELs) have become an important component in data communication systems. Currently commercial lasers operate at 850 nm, where the lasers are made using AlAsGaAs/GaAs semiconductor layers on GaAs substrates. In these lasers the mirrors forming the optical cavity are formed using alternating layers of AlAs and GaAs, with the AlAs/AlGaAs mirror on at least one side of the active region. Adjacent to the active region is an n-type spacer layer, on one side, and a p-type spacer layer, on the other, which inject carriers into the active region when a voltage is applied to the laser. One of the approaches to achieving VCSELs lasing at the important telecommunication wavelengths of 1.3 or 1.55 microns is to fabricate the lasers from materials based on InP substrates.
One problem with this approach is the high free carrier absorption in the p-doped layers of the laser. Free-carrier absorption is the phenomenon whereby an electron or hole within a band absorbs radiation by transferring from a low-energy level to an empty high-energy level. This problem becomes worse as the lasing wavelength increases to the longer wavelengths of 1.3 or 1.55 microns. However, free-carrier absorption is not as significant a problem for short wavelength VCSELs, such as 850 nm VCSELs.
To make matters worse, the poor mobility of the p-type layers results in a non-uniform current injection. Therefore, thick p-type layers may be needed to make uniform current injection. The increased thickness of the p-type layers increases the total optical absorption by the free carrier absorption.
These long-wavelength VSCEL problems can be addressed by using a tunnel junction to replace most of the p-doped layers with n-doped layers, as has been done recently by several groups. Because n-doped layers have a lower free carrier absorption and a higher mobility of carriers than p-doped layers, total optical absorption can be reduced by the replacement of the p-doped layer with a tunnel junction as well as obtaining a uniform current injection. Tunnel junctions or Esaki junctions are well known and can be used, aside from long-wavelength VCSELs, in many other applications, such as solar cells.
Conventional VCSELs without tunnel junctions have one p-n junction which forms the active layers. Carrier injection is the process whereby carriers are injected across a p-n junction, with electrons being injected from the n-layer into the p-layer and holes from the p-layer into the n-layer when an external electric source is applied to the junction.
On the other hand, VCSELs with tunnel junctions require n-p-n junctions. Firstly, the n-p junction is a tunnel junction which has heavily doped n-type and p-type layers. Secondly, the p-n junction is formed by the active layers in the same manner as in conventional VCSELs. When an electrical bias is applied in a forward direction to the p-n junction of the active layers, the same electrical bias appears as a reverse bias for the tunnel junction. As a result, the reverse biased tunnel junction, used in the VCSEL, converts the hole current in the p-doped layers to an electron current in the n-doped layers.
In order to make as near ideal a tunnel junction as possible and to minimize the series resistance of a VCSEL, it is necessary to obtain very high n- and p-doping in the layers forming the tunnel junction.
It is also known that low optical absorption is desired in tunnel junctions for VCSELs. If the optical absorption is low in a VCSEL, the threshold current decreases while the slope efficiency increases. With the increase of slope efficiency, maximum output power of the VCSEL increases.
The optical absorption of the VCSEL consists of optical absorption, scattering loss and others. The optical absorption consists of the free carrier absorption and the bandgap absorption. When the photon energy of wavelength is greater than the bandgap energy, the bandgap absorption, which is typically >4000 cm−1, is dominant compared with the free carrier absorption which ranges about <100 cm−1.
The active layer or region is the layer or region in a semiconductor injection laser or light-emitting diode that provides optical gain. The active region is not really a single layer but multiple layers, each of which can have their own lattice constant. Typically, the active region consists of a multiple quantum wells with compressively strained quantum well layers and often tensile strained barrier layers. The tunnel junction converts incoming electrons into holes that are injected into the active region. Electrons are injected into the active region from the n-side of the active region and holes from the p-side. The recombination of electrons and holes in the active region produces photons, which enable the laser operation.
There can be as many lattice constants as there are layers and these can all be different. One usually tries to keep them the same (lattice matched to the substrate) but sometimes one changes them to introduce strain for added design flexibility. For example, the active region quantum wells are intentionally strained to obtain certain benefits, such as lower threshold and higher gain in lasers. However, the thickness of the strained layer is limited because it creates crystal defects beyond a certain thickness level, resulting in poor performance of devices and poor reliability. Therefore, strained layers should not be used unless it is really needed, as in the active region quantum wells. Feedback resonators, enclosing the spacer layers, to form the optical cavity, are commonly used in VCSELs to obtain lasing.
One technical challenge is that a high probability of tunneling is required in tunnel junctions. Tunneling is an observed effect of the ability of certain atomic particles to pass through a barrier that they cannot pass over because of the required energy, based on a law of quantum mechanics that predicts that the particles have a finite probability for tunneling according to their quantum-mechanical nature. If the tunneling probability increases, the electrical resistance decreases. With the reduction of the electrical resistance, joule heat inside the device decreases. With the reduction of heat, both the maximum output power and temperature performance of the overall device increase.
To increase the tunneling probability, one important way is to ensure that the doping levels of each n-type and p-type layer forming the tunnel junction should be as high as possible. At the same time, a low diffusivity dopant is needed for doping the tunnel junction.
Position control is the reason why a low diffusivity dopant is needed. Typically, the tunnel junction is located at the node of the standing wave of the optical field in the optical cavity. Because optical absorption is proportional to the intensity of optical field, at the node of the standing wave, the absorption is less. Therefore, if the tunnel junction is matched to the position of the node, absorption is less. A low diffusivity dopant is necessary in order to be able to define the position of a tunnel junction. However, if the dopant diffuses everywhere, the tunnel junction no longer coincides with the node of the standing wave and optical absorption increases.
Obtaining a very high p-doping while avoiding diffusion, entails using carbon (C) as the p-dopant, since carbon has a very low diffusion coefficient. Unfortunately, in many materials grown by organometallic chemical vapor deposition (OMCVD), a significant fraction of the carbon acceptors are compensated by hydrogen (H) atoms preventing a high hole concentration. Although this problem can be avoided by growing the materials by molecular beam epitaxy (MBE), the preferred method for high volume manufacture is OMCVD. Therefore, obtaining high p-type doping by OMCVD, the preferred high volume growth technology, has been difficult due to the passivation of acceptors by hydrogen.
It is also known that another way to increase the probability of tunneling is for tunnel junctions to maintain a desirable small difference between the valence band energy (Ev) of the material of the p-type tunnel junction layer and the conduction band energy (Ec) of the material of the n-type tunnel junction layer. To achieve this, it is known that the bandgaps of the materials forming the tunnel junction should be minimized. However, if the bandgap of each individual tunnel junction layer is too low for the lasing wavelength, optical absorption by the bandgap increases. Long-term problems in long-wavelength-VCSELs research remain on how to formulate the optimum compositions of the materials of the tunnel junction layers to meet these criteria and to minimize light absorption in the context of other technical and manufacturing challenges.
Other unknown properties, in particular implementations, include difficulties of doping alternative materials such as AlGaInAs with high p-levels using carbon (C) while minimizing hydrogen (H) passivation. Recently it has been shown that the problem of the passivation by hydrogen is nearly non-existent in C-doped GaAsSb. However, other technical and manufacturing challenges still have to be met. We have found the bandgap of GaAsSb to be too low for use even in 1.55 micron long wavelength VCSELs. We believe that even tensile strained GaAsSb may have too low a bandgap to be suitable for long-wavelength VCSELS, such as at 1.3 micron, due to the bandtails formed in heavily p-doped semiconductors giving rise to excessive absorption at energies below the intrinsic bandgap. In addition to the low bandgap problem, strained layers should still be avoided.
In some approaches of carbon-doped tunnel junctions in long-wavelength VCSELs of some specific material systems such as GaAsSb, the strain and bandgap are interconnected. When the bandgap is not independent of the strain, there is no flexibility to overcome undesired properties of the tunnel junction. Decoupling the strain from the bandgap is well known in some semiconductor devices. However the need or the implementation of decoupling in certain material systems and specific device implementation is not known. In order to decouple, the optimum bandgap value needs to be predetermined so that fabrication resources and time are not wasted.
AlGaAsSb can have a larger bandgap than the lasing wavelength of 1.3 or 1.55 microns. The bandgap of GaAsSb is too narrow and, as a result, it has large optical absorption or loss if GaAsSb is used as part of the tunnel junction. As can be seen in FIG. 4 of the Agilent patent application U.S. 2004/0051113, “InP” is used as a common substrate material at the right side vertical axis. To grow a layer on InP, the lattice constant of the layer has to be matched to the lattice constant of InP. For GaAsSb, about 0.5 fraction of Sb is matched to InP. But, in this case, the bandgap of GaAsSb is about 1.55 um which is not sufficient to prevent excessive absorption in a tunnel junction used in a VCSEL lasing at a wavelength of 1.55 or 1.31 um lasers.
An additional phenomenon needs to be considered, ahead of time, in pre-determining the optimum bandgap. Bandtails are known but their existence in tunnel junction layers used in VCSELs or other devices have not yet been taught. For example, heavy p-doping gives rise to bandtails, which results in absorption at energies lower than the intrinsic bandgap energy of the semiconductor. For p-type layer, we can define an effective bandgap, which is lower than the intrinsic bandgap, below which the absorption can be ignored. The extent of the bandtail and hence the effective bandgap will depend on the doping level Adequate separation has to be maintained between the photon energy of the lasing wavelength and the effective bandgap in order to minimize absorption in the p-layer. The same separation between the photon energy of the lasing wavelength and the intrinsic bandgap of the p-layer of the tunnel junction would give rise to too high an absorption of the laser light.
The detailed nature of how the p-doping and n-doping is used with the particular material systems in the tunnel junction of a long-wavelength VCSEL is a critical issue because the resulting structure must perform multiple functions which include providing a low series resistance, funneling of carriers into the active region, and minimizing the effects of free carrier absorption.
For p-type materials, the bandtails cause a reduction in the energy at which significant absorption takes place (effective bandgap is reduced). In n-type materials, the Moss-Burstein effect (due to band filling) causes a shift of the absorption to higher energy (effective bandgap is increased). Because of the large band filling in n-type materials, the bandtails do not play a role in reducing the effective band gap. For the n-type tunnel junction layer, one can use a material with a smaller intrinsic bandgap than would be the case if the Moss-Burstein shift was absent.
The Moss-Burstein effect can also be present in p-type materials but the band filling is minimal due to the larger mass of holes compared to electrons. Therefore, there is essentially no increase of the energy (which opposes the effect due to bandtailing) at which significant absorption takes place in p-type materials.
Another previously not accounted for effect is the observance that heavy doping causes bandtails to be formed and absorption occurs at lower energies than that of the intrinsic bandgap. These bandtails are not desirable features but something which cannot be avoided and the resultant increased absorption at the lasing wavelength has to be overcome by increasing the intrinsic bandgap of the p-type tunnel junction layer.
Therefore, a simple working tunnel junction that is compatible with a long wavelength VCSEL is desired that overcomes performance-limiting and manufacturing-limiting properties. The manufacturing criterion includes minimal hydrogen passivation of acceptors in an OMCVD reactor. Determining a pre-determined optimum bandgap value, for a decoupled strain and bandgap long-wavelength VCSEL tunnel junction layers that increases the probability of tunneling, without wasting fabrication resources and time is desirable.